0
$\begingroup$

I ran a neural network for 20+ different locations across the United States. At each location I have a list of their predictions in an array. This looks something like this...

loc1 = [2, 0, 1, 1, 2, 0, 0, ... 2]
loc2 = [2, 0, 0, 0, 1, 0, 1, ....1]
...

I am trying to find a way to see what locations predicted the most similarly. i.e. What locations arrays and the most like others.

My first idea was to use k-means clustering. However, when trying to find the appropriate number of clusters using an elbow plot, an elbow doesn't appear and instead it's just a straight line like this... enter image description here

My next idea was to calculate the correlation between all the locations and only looking at correlations that were >0.25. However, none of the locations were greater than 0.25 which makes me think I need to reevaluate how to find locations that predict similarly. Should I take this a different direction?

note: I am coding this all in Python

$\endgroup$

1 Answer 1

0
$\begingroup$

Reducing 0.25 to something like 0.15 would help you find the similar predictions. Similar prediction arrays have high correlation. If you can get a set of arrays with correlation just above 0.15 (for example), then you have found similar predictions. Hope this helps.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.